PDCB concentrations were monitored in 287 households in four southeast Michigan, USA communities: Ann Arbor (AA), a largely suburban and affluent community of 113,934; Ypsilanti (YP), an economically diverse and urbanized town of 19,435, with more commercial activity; Dearborn (DB), population 98,153; and Detroit (DT), population 713,777 (The U.S. 2010 Census). The latter two cities are fairly densely populated industrialized cities that have relatively low household incomes and education levels. Households were recruited using several methods. Random sampling using telephone dialing and snowball recruitment methods garnered 161 households (65, 35 and 61 in AA, YP and DB, respectively) as part of an exposure study (Jia et al., 2008a, b). In DT, as part of a community-based participatory research (CBPR) study on asthma, 126 households were recruited using questionnaires distributed to caregivers at various venues (e.g., schools, and community fairs) aimed at obtaining households in which lived a child (6−12 years old) with symptoms or medication use consistent with persistent asthma (Parker et al., 2008). Informed consent and procedures approved by the University of Michigan Institutional Review Board were followed.

Information on housing, smoking, family characteristics, hobbies and other factors potentially associated with exposure in each household was obtained using questionnaires administered to the head-of-household or child caregiver, and building walkthrough surveys conducted by our technicians. The latter used a standardized form to note each home's characteristics and condition, e.g., type of heating and cooling system, presence of attached garages, and potential emission sources, e.g., incense and room deodorizers. These surveys did not inquire about uses and application rates of PDCB products, as the PDCB measurements were intended to indicate use of these products.

Residences were monitored in at least two seasons. AA and YP residences were visited in summer 2004 and winter 2005; DB residences in fall 2004 and spring and summer 2005. In DT, households entered the study between March, 2009 and February, 2010, and 92% of homes had two or more seasonal visits (11, 17, 87 and 12 homes had 1, 2, 3 and 4 visits, respectively). Monitoring in AA, YP and DB included simultaneous indoor and outdoor sampling. Indoor samplers were deployed in the living room, and outdoor samplers at a location close to the house, e.g., the backyard, at most (90%) homes. Each location used duplicate or triplicate samples. In DT, indoor samplers collected single or duplicate samples in the living room, and duplicate samples in the child's bedroom. Outdoor concentrations were not monitored in DT.

PDCB and other VOCs were measured using passive thermal desorption tube samplers over 3 to 7 day periods in each home (Batterman et al., 2002). Samplers were deployed at breathing height and away from windows, doors, obvious sources of potential contaminants, corners and other potentially stagnant areas, and out of the reach of children. Except for the variance analyses (described later), measurements at different locations in a residence were averaged, and seasonal measurements were averaged to obtain long-term concentration estimates. Thus, each observation is an average of replicates, two rooms for the DT homes, and two or more seasons. Overall, we collected a total of 1,439 valid indoor samples and 478 valid outdoor samples. (Blanks and samples that failed due to sampling or analysis issues are omitted from these counts.)

VOC analyses were performed by an automated thermal desorption-gas chromatography–mass spectrometry (ATD-GC-MS) system (Jia et al., 2006). Given the high PDCB levels occasionally encountered indoors, our 7-point calibration (equivalent to concentrations from 0.2 to 200 µg m⁻³) was extended to 500 µg m⁻³. Linearity remained excellent, for example, the R² was 0.996 and the relative standard deviation of duplicate measurements remained within 9%. Quality assurance (QA) measures included the standard operating protocols, weekly collection and analysis of blanks, regular flow checks, quarterly calibrations, and duplicate or triplicate samples. The method detection limit (MDL) was <0.02 µg m⁻³, and the replicate precision was ≤20%. Non-detects were set to 1/2 MDL.

To understand sources of variation, random effects models were used to apportion the variance of the PDCB concentrations. Following Jia et al., 2011, five variance components were used: seasonal variability (e.g., variance between seasons); measurement uncertainty (variance between replicates); and three types of spatial variability: variances between cities; between residences; and within residences. The latter was estimated only for DT, and specifically between bedrooms and living rooms. The variance analysis was performed for indoor (AA, YP, DB and DT) and outdoor (AA, YP and DB) concentrations. Because the random effects model assumes normality and PDCB concentrations followed log-normal distributions, analyses used log-transformed data.

We also used generalized extreme value (GEV) distributions to fit extreme PDCB and naphthalene concentrations, defined as values above the 90 percentile. Goodness-of-fit was evaluated using Anderson-Darling tests. Bivariate analyses examined PDCB and naphthalene levels in the residences, including contingency tables with quartile groupings for each pollutant, scatterplots and correlations.

Emission rates from PDCB products can be estimated in several ways. As noted earlier, emission rates can be determined using chamber tests. While such tests can be well controlled, results may not reflect in-use or representative emissions. Alternatively, emissions can be calculated for the house by considering the entire residence as a "test chamber:" (1)E=λV(Cin−Cout) where E = emission rate of the house (µg h⁻¹), λ = house air exchange rate (h⁻¹), V = house volume (m³), and Cin and Cout are indoor and outdoor concentrations of PDCB (µg m⁻³), respectively. These calculations used the median house volume V (360 m³) and median air exchange rate λ (0.57 h⁻¹) estimated for the Detroit homes (Du et al., 2011; Batterman et al., 2012b). Outside PDCB concentrations Cout in DT were not measured, thus the median level in AA, YP, and DB was used. Air exchange rates in AA, YP and DB were not measured and, as shown later, Detroit was the community with the most frequent use of PDCB. For the indoor concentration Cin, the 90^th percentile concentration in Detroit was selected to portray residences using PDCB products.

Excess lifetime cancer risks were estimated using the unit risk estimate (URE) from California, i.e., 1.1×10⁻⁵ per µg m⁻³ (California OEHHA, 2009).

The variance components analyses used PROC NESTED in SAS (v9.1.3, SAS Institute, Cary, NC, USA). The GEV analyses used GEV in R (v2.14.1).

PDCB was detected in nearly all (95%) homes. The residence-average PDCB concentrations ranged from detection limits to 2,100 µg m⁻³, and the single highest concentration was 4,200 µg m⁻³ (Table 1). The mean and median indoor PDCB concentrations were 21 and 0.36 µg m⁻³, respectively. Medians varied among the four cities (Kruskal-Wallis test, p<0.01), and were ranked as DT > DB > YP > AA. These measures of central tendency of PDCB concentrations fall within ranges reported in most of the other residential studies, e.g., 1.2 to 69 µg m⁻³ for means, and 0.2 to 1.9 µg m⁻³ for medians (Wallace et al., 1985; Wallace et al., 1988; Kinney et al., 2002; Sax et al., 2004; Sexton et al., 2004; Weisel et al., 2005; ATSDR, 2006; Aronson et al., 2007; Jia et al., 2008c; Logue et al., 2011). An earlier residential study in Detroit in 2006 reported a similar mean concentration (17 µg m⁻³, n=41) (Johnson et al., 2010). Generally comparable levels have been seen in studies using personal monitoring, including the nationally-representative 1999–2000 US National Health and Nutrition Examination Survey (NHANES) study, which found median levels of 1.7 µg m⁻³ (n= 665) among adults (Jia et al., 2008c), and an older national US study conducted around 1988, which reported a median personal level of 2.5 µg m⁻³ (n=1,650) and a median indoor concentration of 1.4 µg m⁻³ (n=2,121; (ATSDR, 2006).

Distributions of indoor PDCB concentrations were strongly right-skewed with skewness coefficients from 4.4 to 8.2 depending on city (Figure 1), and large difference between means and medians (Table 1). Indoor PDCB concentrations were sometimes extremely elevated. Five homes in DT (4%) exceeded the RfC (800 µg m⁻³) in one seasonal visit, although levels in these homes were not high in other seasons tested. As noted, the single highest (7-day) measurement was 4,200 µg m⁻³. When averaged across rooms and seasons, this Detroit house also had the highest concentration (2,100 µg m⁻³). These levels slightly to greatly exceed those previously reported, e.g., 4,000 µg m⁻³ (The Health Effects Institute, 2008), 1,500 µg m⁻³ (m- and p-dichlorobenzene combined) (Wallace et al., 1985), 1,790 µg m⁻³ (99th percentile) (Weisel et al., 2005), and 270 µg m⁻³ (95^th percentile) from a review of 12 residential studies (Logue et al., 2011). While individuals were not asked about their use of deodorizers and repellents, and the walkthrough inspection could only confirm the use of these products in a few obvious cases (e.g., boxes of mothballs were occasionally observed), the sharp rise in the concentration distributions can be used to differentiate the residences where PDCB products are used.

The variance components analysis shows that the variability of indoor concentrations was due mostly to between home-to-home variation (44%), seasonal variation (23%), and city variation (22%) (Table 2). Measurement uncertainty was only 10%. This pattern held among three cities (YP, DB, and DT). In AA, variability was predominantly due to seasonal (53%) and then between-home differences (32%). The between-residence variation likely is attributable to differences in PDCB use patterns, source strengths, and ventilation among the homes, while the seasonal variation likely reflects changes in source strengths and ventilation. PDCB concentrations in bedrooms exceeded levels in living rooms in most (77%) of the Detroit homes (Wilcoxon pair-signed rank test, p<0.01, n=279), and the median concentration in the bedroom (0.72 µg m⁻³) was 32% higher than that in the living room (0.52 µg m⁻³). However, this difference is small compared to home-to-home, seasonal, city-to-city, and measurement variance. Most (87%) of the study homes had forced air heating/cooling systems, and the generally modest PDCB concentration differentials within the homes reflect the well-mixed conditions typically seen in US homes (Batterman et al., 2007; Dodson et al., 2008).

As noted earlier, both PDCB and naphthalene are used as pest repellents and deodorizers. Based on monitoring in the same homes using the same methods, a subset of residences had elevated concentrations of naphthalene (Batterman et al., 2012a). Like the PDCB trend, the Detroit homes had the highest concentrations among the four cities, indicating that Detroit residents use repellents and deodorizers more frequently and more intensively than the families in the other communities. Indoor levels of the two chemicals, plotted in Figure 2, show moderate correlation (Spearman correlation coefficient = 0.31, p<0.01, 2-tailed). A two-way contingency table in which each compound was grouped by quartiles showed 10.4% of residences in the top quartiles of both PDCB and naphthalene levels, and 12.5% in the bottom quartiles of the two chemicals, compared to 6.25% expected if use was uncorrelated (Table 3). Of the study residences, 3% had both PDCB and naphthalene above their 90th percentile concentrations (7.0 and 6.4 µg m⁻³, respectively), compared to the 1% expected if use was uncorrelated.

In comparison to naphthalene, the distribution of PDCB concentrations was much more skewed. This is demonstrated by skewness coefficients of 8.2 for PDCB versus 5.1 for naphthalene, the ratio between peak and median concentrations (19-fold for 90th percentile and median concentrations of PDCB, and 7-fold for naphthalene), and peak concentrations (2,100 µg m⁻³ for the highest seasonal averaged PDCB concentration versus 200 µg m⁻³ for naphthalene). The top deciles of PDCB and naphthalene concentrations fit generalized extreme value (GEV) distributions (Anderson-Darling test, p = 0.86 and 0.76, respectively). The greater skewness of PDCB concentrations likely results from several factors. As noted earlier, PDCB has no other sources besides moth and pest repellents, unlike naphthalene, which has many sources that cause it to be ubiquitous in air (Jia and Batterman, 2010; Batterman et al., 2012a). Second, only a subset of residences uses these products. Third, PDCB has a higher vapor pressure than naphthalene, i.e., a comparably-sized cake or ball will produce a higher concentration, albeit for a briefer period, and thus monitoring may be more likely to encounter peak concentrations associated with recent use of PDCB.

Monitoring results show many households in the four communities use both products, and that Detroit residents use both chemicals most often. These products are marketed somewhat differently, e.g., local stores offer "moth balls" comprised of either naphthalene or PDCB, while most toilet cakes or tablets contain PDCB. However, product labeling does not emphasize the product's composition, and we suspect few users know about the composition, and thus the products are used interchangeably.

Outdoor PDCB levels at residences in AA, YP and DB averaged only 0.09 µg m⁻³ (median of 0.04 µg m⁻³; n=145; Table 1). Based on the variance proportions analysis, measurement variability contributed 55% of the total variance, a high level attributable to concentrations that frequently were near the MDL (~0.02 µg m⁻³). Effects of city, season, and house-to-house variation were similar, each contributing 14−16% of the variance (Table 2). Variance analyses by city again showed that measurement uncertainty was largest (55 to 93%); contributions of seasonal and residence effects differed by city. The small between-house variance indicates that concentrations were relatively homogeneous within a neighborhood. Similar PDCB levels have been seen across industrial, residential and commercial sites in Detroit (Simon et al., 2005) and Yokohama, Japan (Tiwari et al., 2010). Because outdoor samplers were located near the houses, some of our measurements may have reflected air exhausted from houses, or possibly the use or disposal of PDCB near the house.

The mean concentration outside homes in AA, YP and DB was at the low-end of the ranges reported outside homes (0.1 to 5 µg m⁻³), and the median was within the 0.1 to 1.8 µg m⁻³ range reported outside homes (Sexton et al., 2004; Weisel et al., 2005; Aronson et al., 2007; Johnson et al., 2010). Measurements at fixed ambient sites show a slightly larger range, 0.18 to 1.2 µg m⁻³, based on measurements in 11 other US states from 1998 to 2008 (Simon et al., 2005; Hultin et al., 2010). While PDCB concentrations outside of the Detroit homes were not obtained in the present study, levels at 6 Detroit sites in 2000–2001 ranged from 0.2 to 0.3 µg m⁻³ and detection rates were only 16 to 35%; levels at a seventh site (N. Delray, a population-oriented site) averaged 3.5 µg m⁻³ (n=29) (Simon et al., 2005). In 2006–2007, PDCB levels at one of the Detroit sites (Dearborn) decreased to 0.1 µg m⁻³ from 0.2 µg m⁻³ in 2000-1 (Hultin et al., 2010), which was similar to our DB measurements.

The highest outdoor levels (integrated over 3–4 days) ranged from 0.14 (AA) to 1.7 µg m⁻³ (YP). Much higher short-term levels have been reported: 83 µg m⁻³ (24-hr) in Detroit in 2001–02 (Simon et al., 2005); 3.6 µg m⁻³ (integrated 7 days) in Detroit in 2006 (Johnson et al., 2010); 355 µg m⁻³ in New Jersey in 2000 (The Health Effects Institute, 2008); 1.0 to 120 µg m⁻³ in relatively old studies reviewed by ADSTR (ATSDR, 2006); 8.4 µg m⁻³ in the West Louisville Kentucky in 2000 (Simon et al., 2005); and 16 µg m⁻³ in industrial sections of Toronto and Windsor, Canada (Environment Canada, 1993). These levels indicate local sources, or, we suspect, contamination of the sample since PDCB is widely used indoors.

Outdoor PDCB concentrations were far lower than indoor levels in most homes. Thus, outdoor levels will provide only small contributions to the total exposure for most persons. For those homes that do not contain PDCB-emitting products, however, outdoor levels represent a “floor” for indoor concentrations, and thus indoor and outdoor levels will be very similar.

Emission rates of PDCB-based products have been estimated using both chamber and test house experiments. Emission rates of 3.3 to 11 mg h⁻¹ were estimated for individual tablets (7.5–8.4 g each), 3.7 mg h⁻¹ for a block type product (26 g), and 350 mg h⁻¹ for a hanging dispenser (160 g), all at 25 °C using a model for products sold in Japan (Shinohara et al., 2008). Using small chamber tests and air exchange rates from 0.25 to 2.0 h⁻¹, emission rates from moth crystal cakes (surface area 55 cm² each) ranged from 58 to 99 mg h⁻¹ at 23 °C, and 220 to 325 mg h⁻¹ at 35°C (Tichenor et al., 1990). Five moth crystal cakes placed in the closet of a test house gave PDCB emission rates from 381 to 464 mg h⁻¹ over an 11 day period (Aronson et al., 2007). Emission rates for toilet bowl deodorizers (85 g toilet rim block) reported by the US Consumer Products Safety Commission in 1991 averaged 310 and 588 mg h⁻¹ in 13 and 5.5 day-long tests, respectively (Aronson et al., 2007). Several other studies have reported concentrations (but not emission rates) from PDCB deodorizers: median and average personal air concentrations were 340 and 500 µg m⁻³ in 3 day tests of toilet bowl deodorizers (Wallace et al., 1989); concentrations reached 871 µg m⁻³ in residential bathrooms containing deodorizer blocks (Djohan et al., 2007); and spray and liquid deodorizers increased indoor air concentrations to 37 and 25 µg m⁻³, respectively (Wallace et al., 1989). Additionally, six of 26 gel type air fresheners tested in Korea contained PDCB, but emission rate information was not presented (Jo et al., 2008).

Based on eq. (1) and the nominal building parameters, the 90th percentile concentration in Detroit residences (26 µg m⁻³) yields an emission rate of 5.3 mg h⁻¹, which approximately matches emission rates for a single moth "ball" or tablet. For the maximum concentration (2,100 µg m⁻³), the emission rate prediction is 431 mg h⁻¹, which is equivalent to that from one or two block-type products or boxes of moth "crystal cakes." Of course, residences can contain multiple PDCB sources, and the emission rate estimates can be affected by mixing assumptions, the presence of exhaust ventilation, adsorption/desorption (source-sink) effects, and the location of the source, e.g., bathroom, closet, and garment bag. Emission rates also may slowly decline as the mass of the PDCB solid decreases (Shinohara et al., 2008).

The emission rates derived from the Detroit data, which represent real-world conditions, appear reasonable based on the available literature. Emission test results depend on test conditions, e.g., product type, amount, temperature and air exchange rate, and in-use conditions may involve mass-transfer limitations that can significantly lower emission rates compared to those determined using chamber tests. This may be especially important for moth repellents, which are often used in closed spaces with limited air flow, e.g., garment bags, drawers and closets. Use of PDCB-containing products in homes in our study is suggested by PDCB concentrations exceeding 1 to 2 µg m⁻³, a level that correspond to a rapid increase in the PDCB concentration distribution (Figure 1). This threshold suggests PDCB use in approximately 15% (AA) to 35% (DT) of the study homes.

Long-term concentrations of PDCB were far below the current chronic non-cancer RfC (800 µg m⁻³) in nearly all homes (100% of AA, DB and YP homes, 96% of DT homes). Distributions of lifetime excess cancer risks in the four cities predicted using the California URE (1.1×10⁻⁵ per µg m⁻³) are shown in Figure 1 on the right hand axis. For the median and 90th percentile PDCB concentrations across the four cities, the predicted risks are 4.0×10⁻⁶ and 7.7×10⁻⁵, respectively. Depending on the city, 3 to 13% of homes exceeded 9.1 µg m⁻³, a risk level of 10⁻⁴, 0 to 6% of homes exceeded 91 µg m⁻³, a risk level of 10⁻³, and 1% of homes exceeded 910 µg m³, a high risk of 10⁻². Five homes in DT (4%) exceeded 240 µg m⁻³, the Japanese indoor air guideline. Similar results have been shown in several studies, e.g., median risks of PDCB ranged from 10^–4 to 10^–6 (Loh et al., 2007), and 90^th percentile risks for PDCB exceeded 10⁻³ in the RIOPA study (Hun et al., 2009) and in some high schools in New York City and Los Angeles (Sax et al., 2006).

Estimated cancer risks from PDCB at high concentrations exceed those for most other VOCs, e.g., benzene (Jia et al., 2008b; Hun et al., 2009), as well as other pollutant types with the possible exceptions of particulate matter and radon. However, PDCB risks are about an order of magnitude below those from naphthalene (median and 90th percentile risks of 9×10⁻⁵ and 7×10⁻⁴, respectively, calculated using the draft inhalation URE of 1 × 10⁻⁴per µg m⁻³ under consideration by US EPA (US EPA, 2004; Batterman et al., 2012a).

This is the first paper that examines concentrations, exposures and risks of PDCB in indoor and outdoor air in Michigan, and the concentration distributions and variance components analyses demonstrate that both the frequency of PDCB use and usage practices differed among the four cities tested. In particular, PDCB was seen more frequently at higher concentrations in Detroit homes than in the other cities. Of the four communities, Detroit has the lowest household income, lowest educational attainment, poorest health by many indicators, and the highest fraction of African Americans. Discussions with community members indicate that deodorants and repellants (including PDCB and naphthalene) are sometimes used in an “off-label” and inappropriate manner, for example, to mask odors from pesticide applications. Potentially the poorer quality of housing that leads to greater frequency of odor and pest issues, and attitudes and behaviors that lead to greater acceptance of chemical use in homes, are key reasons why Detroit showed the most frequent use and the highest concentrations of PDCB across the four communities. This suggests a need to target policies and tailor educational and awareness efforts to the groups that are most at risk.

The study is also significant in showing that high concentrations of PDCB are not uncommon, the highly skewed nature of the concentration, and the low temporal correlation of concentrations in the households, that is, a high concentration in a home in one season often was not seen in subsequent seasons. This low correlation, along with the variance components analysis, indicate that repeated measurements of PDCB in a large number of homes are needed to estimate long-term averages and exposures that are representative.

PDCB's very skewed concentration distribution produces much higher risks for a subset of residences, e.g., risks exceeded 10⁻³ for 4% of homes, indicating that PDCB exposure can be significant. The current strategy for managing exposures and risks associated with most consumer and industrial products is through right-to-know requirements. PDCB is listed as a carcinogen in California's Proposition 65, in Michigan’s air toxics guidance, in the Safe Drinking Water and Toxic Enforcement Act of 1986, and as a hazardous chemical under the European Union's Regulation on Registration, as examples. In the Registration Eligibility Decision for PDCB, US EPA initially specified a draft inhalation cancer unit risk of 4 × 10⁻⁶ per µg m⁻³, but this was withdrawn due to the lack of evidence regarding carcinogenicity (USEPA, 2008). US EPA then classified PDCB as a low-risk pesticide (USEPA, 2010). Both PDCB and naphthalene are commonly used as deodorizers and pest repellents in residential and other settings. Sometimes these products are applied as insect and animal repellents by placing flakes or tablets on trays or other surfaces in rooms, attics and outdoors in gardens. Such excessive or "off-label" uses can greatly elevate indoor concentrations. While naphthalene has lower sublimation rates and thus lasts longer, the US market has shifted from naphthalene to PDCB due to its lower toxicity. At the same time, fragrances, essential oils, and surfactant-based alternatives have replaced both PDCB and naphthalene as air freshener and toilet/urinal deodorizers. California has restricted the sale of consumer products containing PDCB, including air fresheners and toilet/urinal care products (California Air Resources Board, 2003; New York State, 2004), and New York has restricted the purchase and use of PDCB-containing toilet or urinal deodorizers in school buildings (New York State, 2004). Still, PDCB is widely available and use remains high. The high concentrations found in the present study demonstrate a need to manage PDCB exposures and risks. Possible actions might include sales restrictions, improved labeling and use instructions, consumer education, and promotion of non-toxic alternatives.

This study has several limitations. With respect to the experimental measurements, while homes were measured at least twice, seasonal variability was only partially captured. In AA, YP and DB homes, indoor samples included only the living room, and concentrations in other indoor locations (bathrooms, basement, garage, vehicles, workplaces, etc.) where people might be exposed were not measured. As mentioned, we did not document or ask about specific uses and application rates of PDCB in each home, and thus we cannot determine whether mothballs, room deodorizers, toilet deodorizers, or some other source is responsible for the high concentrations found. The study households may not be representative of those in Michigan or the US more broadly. While several sources associated with high PDCB concentrations were identified, some of the analyses assumed building characteristics and used PDCB emission rates from the literature. Lastly, the health risks presented are screening level estimates. They were calculated using indoor air measurements, which do not represent life time exposures, and they may be biased upwards because they do not account for the amount of time that individuals spend in homes, temporal variation, other exposure sources, and other microenvironments.